Review: Scythe Ninja Revision B
01.25.07 - 07:46pm
If you were to take a peek under the hood of most of the boxes shipping from Dell and other OEM companies you would think that heatsink technology has been rather stagnant over the last 2 years. It is a rather sad fact that the stock heatsink for both Intel and AMD processors are so vastly underpowered but they are also massproduced chunks of aluminum that cut a fine line between performance and price. The aftermarket heatsink market on the otherhand has a broad range of heatsinks ranging from near-OEM models up to multi-fin multi-pipe monstrosities generating enough thrust to help launch a satellite into orbit (slight exaggeration..). Today I’m going to perform a review upon my personal Scythe Ninja Rev. B, one of the top heatsinks on the market at this time.
First Impression
At first glance the Scythe Ninja makes me think of vegetable slicers that I have seen in some kitchens. Twenty-three aluminum fins stacked horizontally compose the main heat-dispersion area with six u-bent heatpipes transferring the heat from the processor to the fins. The boxy design of the Ninja gives the heatsink roughly half a meter of surface area over all the fins along with a smaller finned heatsink sitting sandwiched between the baseplate and the aluminum fins. Measuring in at 110mm by 110mm by 150mm and 665 grams, some consideration needs to be taken as to where you will utilize this heatsink. Most SFF cases won’t be capable of housing this heatsink and even some midtower cases may not be wide enough for the heatsink to fit. When trying to make a decision, remember that if your case has a side-fan to factor that into the dimensions as a few users have complained about the heatsink being too tall. On a select few motherboards the socket has been positioned too far towards the top of the motherboard and the heatsink may collide with the powersupply if the case doesn’t have enough clearance between the board and the powersupply. To test for compatibility I suggest measuring 6cm from the center of your socket and checking that the heatsink will clear your powersupply. The actual heatpipes may be in close proximity to the Northbridge depending on the motherboard, on my DS3 there is less than 1mm of clearance however neither heatsink makes contact.
Applications
Due to the fin density of the Ninja it can be used in two instances, passive or active cooling. Passive cooling a processor is relatively unheard-of in this day and age however it is possible if proper preparation is taken. The heatsink ships with a 120mm fan along with fan brackets and from what I can tell the fan is of rather high quality with very low acoustical values. The Ninja makes for a great general purpose heatsink by offering performance that rivals some low-quality watercooling loops while retaining the maintenance requirements of an air heatsink. With the stock fan the heatsink does a great job at dissipating the heat produced by a Core2Duo and it has been reported of being capable of handling Core2Quads are moderate clockspeeds. In silent environments the Ninja would make a great heatsink especially in an HTPC case that has enough space to clear the heatsink.
Test Platform and Procedure
I will be performing the temperature tests on my current test bed. I had planned on running these tests on an E6600 also but the chip has failed to show at my door. For a general comparison the stock Intel heatsink will be tested with the fan controller set to fullspeed. The Intel heatsink spun up to ~2500 RPM while the Ninja hovered around 1100 RPM through the testing. In both cases a second 120mm fan was place over the memory to actively cool the memory modules and provide some extra cooling to the Northbridge. Active memory cooling isn’t necessary for stockspeeds however once memory voltages go above the JEDEC 1.8 volts the temperatures climb quickly. The Ninja was positioned such that the aluminum heatsink near the base was aligned with the fan and the entire assembly was blowing the top of the board towards the Northbridge/Video card. These tests were performed outside of a case on my benching station with the ambient temperature ranging between 19 C and 20 C. The core temperatures were measured with the program Core Temp and were the average between the two cores. Idle temperatures were taken after a fresh boot and five minutes had passed since the Windows GUI had loaded. Load temperatures were taken after 25 minutes of running 100% load. SuperPi 32M was the chosen application for loadtesting with 4 instances being looped to ensure 100% processor load. Besides the 4 instances of SuperPi 32M and 1 instance of CoreTemp, the system was running the processes that are loaded on a fresh installation of Windows XP Professional Service Pack 1.
- Processor: E6400 2.13GHz @ 1.25 V
- Motherboard: Gigabyte DS3 Rev1 w/ C2 Northbridge
- Memory: 2 x 1GB Crucial 10th Anniversary DDR2-667 3-3-3 modules @ 1.8 V
- Video: EVGA 7900GTO @ 650 core 660 memory
- HDD: Seagate 7200.10 300GB drive formated with NTFS, S.M.A.R.T. enabled
- PSU: OCZ GameXtream 750 Watt
- Case: Modified Ikea Table
Prior to performing the actual load-testing I cycled through my various known stable voltage levels and trimmed as much core voltage off as possible while retaining stability. Starting with 1.25 vcore and scaling up to 1.425 vcore let the processor scale from 2133MHz to a final 3600MHz. 3600MHz is not the peak operating frequency of this processor as it has scaled past 4300MHz however 1.425 vcore is the maximum I will push through processor on aircooling. One unusual item that I noticed with the Core2Duo voltage/frequency scaling was that it may take 1.425 volts on air for the system to be stable on air but when under phase it would happily scale as high as 3832MHz with 1.425 volts. Memory and Northbridge voltages were adjusted to maintain stability however they will not effect CPU temperatures so those settings were omitted. All forms of thermal management were disabled for the tests as they tended to interfere with the overclocked stability of the system, causing it to crash. For each individual run the idle temperature was taken, the processor was loaded and then the load temperature was taken, and then the system was shutdown and a 30cm deskfan was placed against the board to thoroughly dissipate any latent heat.
Results
I was very impressed with the Ninja heatsink when comparing it to the stock heatsink. To say the Ninja is superior would be a supreme understatement, the Ninja obliterated the stock heatsink’s results. I believe the heatpipes in the Ninja have an optimal operating temperature that lay somewhere between 40 and 50 celsius as the temperature deltas between idle and load remained very consistent at these temperatures.
From 2133MHz to 2666MHz the Ninja maintains a 100% load temperature that is lower than the Intel idle temperatures. The jump to 3200MHz was a large jump in vcore and the temperatures reflect that however the incremental increases up to 3600MHz show the performance of the Ninja level off as the heatpipes become loaded. While I didn’t perform any tests above 1.425 volts, I know from experience that the system will run up to about 3800MHz at 1.475 and the temperatures will begin to crest past 65 Celsius so the heatsink is surely capable of running at higher clockspeeds. With a quadcore processor I would expect the performance to diminish considerably due to the 4 cores dumping double the heat of the 2 core E6400. My rough estimate would put a Q6600 at 65 celsius @ 3600 MHz with the possibility of seeing 70 celsius not unreasonable. The Scythe Ninja has an immense cooling capacity but even it has a limit.
Final Thoughts
The Scythe Ninja does an excellent job at cooling what I would consider to be a large amount of heat. My only problem with the heatsink would be physically mounting it to the motherboard as it did a fantastic job of slicing one of my fingers. This problem could be avoided by wearing gloves or taking your time but do be aware that the possibility exists of walking away with a few cuts. The Ninja is priced decently in the upper $30-lower $40 bracket and provides a cheaper alternative to the Scythe Infinity and Tuniq Tower 120. I believe the performance of the Ninja lies slightly lower than the other two heatsinks but we are talking in the range of 2-3 Celsius which is rather small compared to the 10-17 Celsius temperature delta between the stock heatsink and the Ninja. Scythe has made an excellent heatsink and I eagerly look forward to their next wave of products.
Speak Your Mind